OpenAI, the AI giant behind ChatGPT, plans to make its own AI chips using TSMC’s 1.6nm technology, it said. TechNode Quoting the Chinese source Economic Daily News.
Why is it important?This move would revolutionize OpenAI, which would design its own chips, like Apple does, allowing it to have greater control and giving its chips a specific specialization toward its interests, in addition to reducing costs (presumably) and increasing its capacity compared to Google or Microsoft.
It would also be a shock to the cornucopia that NVIDIA has become, the main supplier of specialized AI chips for almost the entire industry.
In detail.
- OpenAI has reserved production capacity from TSMC to use its 1.6nm process, known as A16.
- This A16 node represents a leap forward from TSMC’s current N4 and N3, although it is still in development.
- Production is expected to begin in 2026, a year earlier than Intel and Samsung had planned.
Behind the scenesThere are rumors that Apple could be a collaborative partner of OpenAI in this project. These are not empty rumors: ChatGPT is part of Apple’s new operating systems thanks to its native integration, and a few days ago we learned that Apple is one of those interested in entering the next round of financing of OpenAI.
Apple also has a long history of designing its own chips, using TSMC for manufacturing, but it would not be a competitor (at least for now) for OpenAI, but rather an ally.
The context. OpenAI is spending astronomical amounts on NVIDIA servers to run, primarily, ChatGPT, a product that still has a more than generous free offering. Each query is estimated to cost about 4 cents.
If ChatGPT usage reaches one-tenth of Google search volume, OpenAI is said to have an annual turnover of around $16 billion in chips. It is logical that its next move will be to cut costs in order to get closer to profitability.
Between the linesThe potential collaboration with Apple suggests that OpenAI is looking for powerful allies to compete against those launching similar products and services. Apple Intelligence has a somewhat different approach today, more focused on local execution and horizontal integration, rather than being a standalone service.
Precisely because it is independent, it would be more difficult for OpenAI to remain in a scenario in which excessive investments are needed to stay in the race, as Antonio Ortiz pointed out a few days ago in Error500.
At Xataka | TSMC is going to bet on glass substrates at the request of NVIDIA. In this race, Intel is in first position
Featured Image | TSMC
Add Comment